onel.bsky.social
Founder askmanu, helping developers get answers
14 posts
16 followers
74 following
Regular Contributor
comment in response to
post
Is that linkedin?
comment in response to
post
Not the main reason, but yeah, plenty of good alternatives and less drama associated with them. Also, I don't even know if Grok has an API platform
comment in response to
post
The second point got me thinking that specialized models might be something we are overlooking now because of the hype of AGI.
Being a video from a science conference, of course it's 10h long.
comment in response to
post
👉 this whole field is really held back of this aspiration of building this super-robot that helps you with all your tasks
comment in response to
post
Some ideas from the talk:
👉 AGI does not have all the context: "what's in my mind right now is the context of what I'm going to do next today in life. It cannot know that. And if it doesn't know that, it can't give me good advice."
comment in response to
post
This is great to see, because not only the results are better than using O1 which is SOTA at this point but the cost is also 𝟭𝟰𝘅 𝗹𝗲𝘀𝘀.
comment in response to
post
This interesting blogpost from the Aider project, explored the idea of combining the two to create the top results on their polyglot benchmark.
They used R1 to think about a task and create a plan and then used Claude Sonnet for the actual implementation.
comment in response to
post
Unfortunately there are still some reasons like cost or ease of use that make closed model a good choice. Inertia is also big